Engines of power: Electricity, AI, and general-purpose, military transformations

Jeffrey Ding, Allan Dafoe

Introduction

In contemporary defense discourse, Artificial Intelligence (AI) is frequently likened to "the new electricity." Yet, despite this analogy, many military scholars continue to analyze AI through the lens of "narrow" technologies—discrete breakthroughs like nuclear weapons or aircraft carriers that offer immediate, localized advantages. This paper challenges that perspective, arguing that AI is a General-Purpose Technology (GPT) that triggers a "General-Purpose Military Transformation" (GMT). Unlike narrow innovations, GMTs do not necessarily level the playing field; instead, they often reinforce the power of established states with deep civilian industrial bases.

The GMT Framework

The authors derive the GMT concept from economic literature, identifying three defining characteristics that distinguish it from narrow technological shifts: breadth of impact, significant time delays, and indirect productivity spillovers.

Comparison of Narrow vs. GPT Innovation Pathways

As illustrated in the comparison above, narrow innovations follow a relatively linear path from innovation to specific applications and strategic impacts. In contrast, a GPT creates a complex web of complementary innovations and diverse applications. This complexity leads to a "gestation period"—often spanning decades—as organizations restructure to fully exploit the technology. Furthermore, the most significant gains often stem from "boring" improvements in logistics, decision support, and manufacturing readiness rather than a single "silver bullet" weapon system.

Historical Evidence: Electricity vs. Submarines

To test this theory, the researchers compared the military adoption of electricity (a GPT) with submarine technology (a narrow, non-GPT) during the late 19th and early 20th centuries. Using data from the CHAT dataset and naval power records, they examined the divergence between Great Britain and Russia.

The findings reveal a stark contrast. In the case of submarines, Russia—despite its industrial laggard status—was able to keep pace with Britain because the technology was narrow and did not require a massive civilian infrastructure. However, in electrification, Britain maintained a massive lead. While Russia produced world-class "star" researchers who pioneered electronic countermeasures, they failed to scale these innovations because their civilian industrial base lacked the necessary talent, standards, and infrastructure. This "civilian dependency" meant that the military could not sustain a GMT without a robust commercial ecosystem to tap into.

Implications for the AI Era

Applying these historical insights to the current US-China military balance, the authors suggest that the "levelling myth"—the idea that AI's commercial nature allows laggards to "leapfrog" leaders—is likely incorrect.

  • The Gestation Phase: Just as American manufacturing didn't see the full productivity gains of electricity until forty years after the dynamo's invention, AI is likely in a similar gestation phase. Rapid transformations predicted within the next decade may underestimate the structural and organizational delays inherent in GPTs.
  • Reinforcing Power: Because AI exploitation depends heavily on a deep civilian ecosystem (talent pools, data infrastructure, and hardware manufacturing), it is more likely to reinforce the power of states that already lead in these areas.
  • Focus on the "Boring": Strategic advantage in the AI era may come less from "killer robots" and more from the incremental, pervasive upgrades to the "engines of power"—the logistical and industrial foundations of military effectiveness.

Conclusion

The transition to AI-enabled warfare is not a sprint toward a single breakthrough, but a marathon of organizational and industrial adaptation. By understanding AI as a GPT, policymakers can move past the search for silver bullets and focus on the long-term structural advantages provided by a robust, integrated civilian-military technological base.

Summary updated

Abstract

Abstract Major theories of military innovation focus on relatively narrow technological developments, such as nuclear weapons or aircraft carriers. Arguably the most profound military implications of technological change, however, come from more fundamental advances arising from ‘general-purpose technologies’ (GPTs), such as the steam engine, electricity, and the computer. Building from scholarship on GPTs and economic growth, we argue that the effects of GPTs on military effectiveness are broad, delayed, and shaped by indirect productivity spillovers. We label this impact pathway a ‘general-purpose military transformation’ (GMT). Contrary to studies that predict GPTs will rapidly diffuse to militaries around the world and narrow gaps in capabilities, we show that GMTs can reinforce existing balances if leading militaries have stronger linkages to a robust industrial base in the GPT than challengers. Evidence from electricity's impact on military affairs, covering the late nineteenth and early twentieth centuries, supports our propositions about GMTs. To probe the explanatory value of our theory and account for alternative interpretations, we compare findings from the electricity case to the military impacts of submarine technology, a non-GPT that emerged in the same period. Finally, we apply our findings to contemporary debates about artificial intelligence, which could plausibly cause a profound GMT.
BibTeX
@inproceedings{ding2023engines,
  author = {Jeffrey Ding and Allan Dafoe},
  title = {{Engines of power: Electricity, AI, and general-purpose, military transformations}},
  booktitle = {European},
  year = {2023},
  doi = {10.1017/eis.2023.1},
}
Figure